TW583406B - Method for analyzing final test parameters - Google Patents

Abstract

A method for analyzing final test parameters, which is utilized for analyzing a plurality of lots of products. Each of the lots has a lot number, and each wafer of these lots has been tested according to a plurality of final test items. Moreover, the final test items, the final test parameters, and the correlations between the final test items and a packaging station, which includes a plurality of packaging machines, are stored in a database. The method includes the following steps: retrieving the final test parameters of each lot of products from the database; comparing the final test parameters to select out a representative final test parameter and a corresponding representative final test item; determining whether the representative final test item is correlated with the packaging machine item or not; when the representative final test item is correlated with the packaging machine item, dividing the lots into two or more groups, including a first qualified group and a first failed group, based on the representative final test item; searching the packaging machines through which the qualified group has passed; searching the packaging machines through which the failed group has passed; and estimating the packaging machines in which the probability that the failed group has passed through is greater than the probability that the qualified group has passed through.

Description

583406 V. Description of the invention (1) '), [Technical field to which the invention belongs] The present invention relates to a method for analyzing process parameters, and more particularly to a method for analyzing test parameters after packaging. I have waited many times; also Taiwan, in order to ensure the title and quality data (I η-1 data, test) after packaging, it is produced at this time (pin) to the cooked part. [Previous technology] Semiconductor manufacturing technology Manufacturing processes, such as lithography, which has been used in semiconductor manufacturing and many tedious procedures. The machine is operating normally, and maintenance and repair of the machine can meet customer needs. Generally speaking, half-hand analysis should be discussed, including ine QC data, missing samples σσ, then the § formula (sample data and post-package test data are used in the production technology of crystal semiconductor components. Please know that the technicians will conduct electrical tests for various post-package tests ^. 着 '在In Step 102, 'To complete a semiconductor product, it is usually necessary to use a large number of machines in the process, etching process, and ion implantation process. Therefore, those who are familiar with this technology are committed to maintaining or improving product yield. Detect and confirm the question 'In order to speed up the production of semiconductor products. Conductor process problems can be determined from the following several process parameter data, online quality test trap detection ( defect inspection (test) test) data, wafer test (wafe; r field (final test) data. Among them, after the circle is cut and the packaging process is performed, the inspection value obtained by the product test is performed. As shown in FIG. 1, the first step is to perform 1 〇1, the test of each semiconductor element after the packaging process, such as the pins of semiconductor components, a skilled person will observe each semiconductor

Page 8 583406

The results of various post-package test items of the component. In order to find out the process questions selected by the tester after packaging, such as the newest set of 4 new questions, the available machine steps are based on the experience of skilled technicians and from step 102. The results of the post-encapsulation test of the different production ports can be used to determine the station type, such as the packaging process, or the station type of other test data, the quality inspection station line, the sample test station, etc. °

After that, in step 104, a skilled technician checks each of the stations in the station in step 3 in order to find the abnormal station or repeat the tests in the test station in step 103. Set value. For example, if a person skilled in the art judges that a certain pin of the semiconductor device is questionable + to search for the process station for this packaging process, and check for abnormalities such as wire bonding machines, sealing (m ο 1 ding) ) Machines, etc .; In addition, if a skilled technician judges that the semiconductor device 4 is related to the previous manufacturing process, it may be that there is a problem with a previous online testing station, resulting in the inability to effectively detect the defective product. Therefore, those skilled in the art will modify various preset specifications of this online testing station to avoid the same problem in subsequent products. 、 ° However, because in the conventional technology, artificial analysis is used to determine the analysis result (step 103) and the revised value (step 丨 〇4), the accuracy and credibility of the final analysis result To be discussed; coupled with the frequent change of people in the semiconductor manufacturing industry, it is difficult to pass on the experience between engineers in the early and late stages, and each engineer has limited capabilities and cannot take into account the operating status of the plant 2 machine station. Therefore, when testing semiconductor products When the result is abnormal, the engineer may not have enough experience to quickly and correctly determine

583406

Which part of the question is related research, even reducing the efficiency of the process and the production situation to improve the yield rate problem, so it may be necessary to make a wrong judgment to increase production costs, and it takes a lot of time to break, so-come, not only I ca n’t improve the post-package test of Risheng & what is the question of the current semi-conductor? Therefore, how to provide a method to quickly and correctly determine the problem when the semiconductor product data is abnormal, and correctly Corrects one of the important issues of the analytical method body manufacturing technology of preset specifications. (Three), [the content of the conductor in accordance with the clear content of the above issue] The above-mentioned subject, the purpose of the present invention is that the test data after the package is broken, which is the problem of the link after the package analysis of the online testing of the quality inspection product of the present invention method. Common analysis data of this invention. The reason is that the analysis method is that the supplies are complex and complex. Another purpose is to provide a test or sample test result to make the standard (the characteristic of the control spec is to analyze the poor products with better product methods to achieve the above purpose. According to the present invention, it is made by analyzing multiple batches with several machines, and the inspection of each post-packaging test item is to provide a method for analyzing the test parameters quickly and correctly in the semi-permanent manner. (Modify the online quality inspection or sample) after packaging test parameters are divided into control groups, and the statistics and test data and machine package after the test parameters of the batch number of products, each batch of each batch of products After circular testing to produce multiple packages

583406 V. Description of the invention (4) ----- Test parameter values' These post-package test items, the post-package test parameters ί Yu Ϊ :: Related to the post-package test items-packaging process station In the shell material library, the method includes the following steps: • Search for the tilt seat to obtain the test parameter values after packaging for each batch of products; compare after mounting: the parameter values are selected-a representative post-package test Parameter two = 、 " Ϊ: 性 general post-package test items; judging a representative post-package comparison of a plurality of batches of representative post-package test items ^ number = preset specifications in order to make the representative Those who are larger than the preset specifications after encapsulation are classified as a first unqualified product, and those whose representative parameter values are less than the preset specifications are classified as a first. : Ge product group, search for the machine that each batch of products of the first qualified product group passes at the packaging process station; search for the machine that each batch of products of the first unqualified product group passes at the 4 process station; and judge the first The probability of passing the unqualified product group is higher than the first Product group through the chances of the machine. In addition, each wafer in each batch of products has undergone double testing of the online quality inspection project and a sample test project related to the 2 test items after packaging to generate an online quality inspection parameter value and a sample test parameter value. The # material library further stores these data, and according to the post-package testing of the present invention, the ^ analysis method uses statistical analysis methods to analyze the results of the post-package test and online mouth quality inspection or sample testing in order to correct the online quality inspection. Or regulatory standards for sample testing. As mentioned above, the method for analyzing the test parameters after packaging according to the present invention is to use a better number of products as a control group, and use statistical and common analysis methods.

Page 11 V. Description of the invention (5) f = Various test data and machine data of the product with poor analysis, so that it can be quickly and correctly when the test data of the f-body product is abnormal after packaging. Which link has a problem so that you can correctly judge ▲ if there is a conclusion ^: The process station 'and then find the abnormal machine' and can be based on the analysis: ancient 1 to modify the online quality inspection or sample test control standards, so it can produce Reduce artificial judgment errors to improve process efficiency, reduce production costs, and improve online production in a timely manner to improve yield (four), [implementation method] Post-installation testing: ί 4 :::: 4 The seal symbol of the preferred embodiment of the present invention will be described. '' Where the same components will be shown with the same reference, as shown in FIG. 2, the deer _ test parameter analysis method is not a packaged machine according to a preferred embodiment of the present invention. , _ Cheng map. This embodiment analyzes and finds the problem as shown in Figure 2. First, in the + case of the test parameters after encapsulation, eight ν and step 20 1, the test parameter values after encapsulation are preferably implemented according to the present invention. A plurality of parameter values of several batches of products, and then one is selected. The 202 is to compare the post-package test parameters to one of the representative packages. The post-package test parameter values and relative n batch products are compared. The seal is connected to the project ^ project. For example, 'assuming there are three types of projects, all, ^ 忒 items have A items, B items, and c. The test parameter values after the break are shown in the following table. 583406 V. Description of the invention (6) c item 7% g M, The percentages indicated by the staff members are the batch numbers in each seal wearing #, 4 clocks, and sheep. In this step, the representative ^ &, "", which is the highest average, A project. In other words, the post-packaging test items after assembly are strangling (k i 11 ed) after sealing /, and the representative item ′ is the item with the highest average failure rate among several batches of products. After installation testing In this embodiment, each lot of product is “lot surface ber”, and each lot of products includes 25 wafers. One lot is a number: a; system; Taiwan, the last batch of a-a clothing test test participation, then, step 2.3 is a representative judgment: value. Whether it is related to the packaging process station. Generally speaking, the test items are @ will be compared with some specific process machines. The test items are stored in a credit after packaging. To avoid two = the example is to test the related post-test items. After a plurality of packages are generated, the J-channel package is then followed by step 203, and the scoring is performed. &Quot; ^ ^ page 13 583406 V. Description of the invention (7) q. When it is judged in step 20 3 that this representative post-package test item is not related to the package, this method will start from other related processes to perform the weight distribution (as shown in Figure 3), such as online quality testing or sample testing. Step 203 judges this representative post-package test item and the packaging process two, and then proceeds to step 204, which is based on representative 'packages, personnel, and several batches of products are divided into at least one first qualified product group And a Ϊ: Α does not top the qualified product group. For example, t, this step is to analyze and compare each batch of production: (such as whether the value of the parameter (disqualification rate) of 2 2 is greater than the preset rule of item A 〇U / o}; otherwise, this batch of products is classified as group A (Products in the first qualified product group j include, for example, batch numbers 1, 2, 3, 4, and 5 (as in No. 5, if so, this batch of products is classified as group B (the first unqualified product ... for example Including the batch number is used to calculate the "probability that all products in this product will pass through the multiple stations of this cover; in general, a two-line system includes several stations'" such as Ε 1, Ε 2, Ε 3 .... In addition, Ϊ1 封装 该等 of these packaging process stations ^ Ϊ⑽1ΓΪ⑽ In step 2 ° 9, use the common analysis method to find out the group 产 product. The probability of two k is 咼 in the group Α product passing probability Therefore, these Group B products obtained in step 209 have a high probability of passing through the machine.

Page 583406

V. Description of the invention (8) A packaging machine which can be analyzed by the analysis method of the post-package test parameter analysis method of the preferred embodiment of the invention. / In addition, please refer to FIG. 3, when it is judged in step 203 that the representative post-package test items are not related to the packaging process station type, then step 3 ^ 1 is performed, which is based on the representative package The post-test project divided several batches of products into a second qualified product group and a second unqualified product group. In this embodiment, 'this step is to analyze and compare whether the parameter value (disqualification rate) of item A of each batch of products is greater than the preset specification of item A (such as 20%). If otherwise, " Classified in the second qualified product group; if so, this batch of products is classified in the second unqualified product group. Then 'step 3 2 2 searches the database for the first eight cells. In the group and the second unqualified product group, the online quality inspection items or sample test items and their parameters related to item A (with =: =; test items). In this embodiment, this step is to select an online quality detection item related to item A (as shown in step 303). Then, steps 304 and 305 are used to statistically analyze the second incompetence. Production. Is there a difference between the search results of the search results of the first qualified product group and the search results of the first qualified product group? In the present embodiment, step 3 0 4 is the first analysis of the parameter values of the online quality inspection items searched at step 3 3. To calculate its mean and variation; then step 30 compares the second

P. 15 583406

V. Description of the invention (9) The online products of the unqualified product group have been prepared from the first paper and the average number and variation number of the detection parameter values are: Find out whether there is a difference between the ancient 圼 s J double Han literate soldiers ^, ¾-^ s ^ 疋. The judgment conclusion of the right step 3 0 5 ..., one ... difference, it means that the cause of the unacceptable rate of the A item is not too high. I. 丨 疋 Ώ & on the ankle. Mouth testing items (such as step Λ not Λ ^ analysis will be stopped at this time; if the result of step 305 shows that there is a difference between the day guards, it means that the A rate of the project failed. The reason for the high value is related to the online quality inspection item selected in step 302 (as shown in step 3003), so the next step is step 3 06. In steps 3 06 to 31i, the main analysis is Find out what caused the problem with the online quality inspection project, and try to provide a method to modify the control spec of the online quality inspection project. In step 3 06, according to the online quality inspection project selected according to step 3 2 (As shown in step 303), the number of out of spec parts of the first unqualified product group and the out of spec parts of the second qualified product group are calculated statistically, and both The second unqualified product The ratio of the number of product batches of the group to the second qualified product group. In this embodiment, the preset specifications of the online quality testing items are a certain range, which has a preset spec upper limit (UP spec limit) and a preset spec lower limit. (Low spec limit). Therefore, this step is to calculate the second

583406

V. Description of the Invention (ίο) The ratio of the number of batches of products in the qualified product group exceeding the preset specification limit to the total number of product batches in the second unqualified product group pa L, and the ratio of the second qualified product group exceeding the preset specification limit The number of product batches in the total production of the second qualified product group: the proportion of product batches PaH; In addition, this step also calculates that the number of product batches in the second unqualified product group that is lower than the lower limit of the preset specification accounts for the second unqualified product The ratio of the total number of product batches in the group, PbL, to the ratio of the number of product batches in the second qualified product group below the preset lower limit to the total number of product batches in the second qualified product group.

PbH. Next, step 307 compares PaL and PaH, and PbL and PbH, respectively, in order to determine whether there is a portion in the second qualified product group and a second unqualified product group that exceeds the upper limit of the preset specification and a portion that is lower than the lower limit of the preset specification. difference. If there is a difference in the result of step 30 7 judgment, proceed to step 3 08, which is judged to cause the representative post-package test parameter value (the parameter value that stifles the post-package test item) is too high. 2 The deviation of the test results of the online quality inspection items (as shown in step 303) found. In the present embodiment, step 308 can use a CDF chart to output the test results of the online quality detection items (shown in step 3 03) searched in step 3 02. Therefore, the engineer can refer to the CDF chart to calibrate Correct the specifications of online quality inspection. In addition, if the result of the judgment in step 3 7 is no difference, go to step

Page 583406 5. Description of the invention (11) 3 0 9 is a statistical analysis of the second unqualified product group (ou t 〇fnarrowedspec) and the first The number of non-conforming product groups of the second qualified product group is reduced, and the proportion of the two is the product batches of the second unqualified product group and the second personality product group. In this embodiment, the reduced specification of the online quality measurement item is a certain range, which has an up narrowed spec limit and a reduced specification lower limit (10 w narrowed spec 1 imit). It should be noted that the range of the above-mentioned preset specifications is usually six times the standard deviation of the online quality inspection process, while reducing the standard deviation of the difference between the upper limit of the specification and the upper limit of the preset specification by one, and reducing the difference between the lower limit of the specification and the upper limit of the preset specification Double the standard deviation, so narrowing down the range of the specification is usually four times this standard deviation. Therefore, 'Step 3 0 9 calculates the ratio of the number of product batches in the second unqualified product group that exceeds the upper limit of the reduced specification to the total number of product batches in the second unqualified product group, PaL', and exceeds the reduction in the second qualified product group. The ratio of the number of product batches with the upper specification limit to the total product batches of the second qualified product group paH; In addition, this step also counts the number of product batches that are lower than the lower limit of the reduced specification in the second unqualified product group, accounting for the second Proportion of the total product batches of the non-conforming product group PbL * _The ratio of the number of product batches below the lower specification limit in the first qualified product group to the total product batches of the second qualified product group.

583406 Description of the invention (12) Next, step 3 1 〇 from the eight, cut, in order to determine the i-th series respectively PaL, and PaH ', and pbL, and those that exceed the upper limit of the reduced specifications: qualified product group and the second unqualified product There are differences in the group. If step 307, points, and the part below the lower limit of the reduced specification: whether the project failure rate is too large: if the result of the judgment is no difference, it means that the quality inspection item A is caused (eg, the reason is not step 3 0) 2 The selected online product is judged by the social fruit in step 310 (shown in step 303), and the analysis is stopped at this time; if there is a difference in the representative seal #, step 311 is performed, which determines the value of the parameter). The test results of high, post-test parameter values (kill the test items of the post-package test items (as shown in the example of online quality inspection in step 3 02, as shown in step 2 02) are offset by the test results. Cause: In this case, we can also use a CDF chart to output the results of step 302. Therefore, the worker knows the specifications of the test knot detection of the item (as shown in step 303). i ί can refer to this CDF chart to calibrate and correct the online quality. In addition, in this preferred embodiment of I Β > σ, I, θ \ Ming found, step 3 2 2 3 3 11 8 Yupin test items, at this time, in the above steps 3 0 3 ~ the test items (all @analytic comparators ie The online quality testing project was changed to the sample tester (not shown in Figure 1), and the analysis results can also provide the specification data of the engineering t-rate correction sample test. The above-mentioned, because of the package according to the present invention Post-test parameter analysis method

Page 19 583406 V. Description of the invention (13)

: It is based on better batches of products as a control, board, and a common sealer with problematic performance, or online quality inspection or sample test data of 2 poor products using statistical analysis methods, so it is the second semiconductor product. Alienation of the test data after encapsulation ^ In determining which link is the problem, &, ', quickly and correctly answer the process station, and then find the abnormal ^ ^ i, and determine the results to correct Online quality inspection or analysis can effectively reduce human judgment errors according to analysis = two control standards, so the production cost and timely improvement of the efficiency of the online / question process, and the reduction of the above is only exemplary, and the 1 form To improve yield. The spirit and scope of the present invention are: Any departure shall be included in the scope of the attached patent application. Ding's equivalent modification or change, both

Page 583406 Brief description of the drawings (five), [Simplified description of the drawings] Fig. 1 is a flowchart showing a conventional method for analyzing test parameters after packaging, and Fig. 2 is a flowchart showing a preferred method according to the present invention. The flow of the post-package test parameter analysis method of the embodiment; and FIG. 3 is a flowchart showing a flow continuing the flowchart shown in FIG. 3. Description of component symbols: 101-104 The flow of the analysis method of the test parameter analysis after packaging 201- 2 0 9 The flow of the analysis method of the test parameter analysis after packaging in the preferred embodiment of the present invention 301 ~ 311 Continue the flow of step 203

Page 21

Claims (1)

583406 VI. Application for Patent Scope 1. A post-package test parameter analysis method has a batch of products, which are produced in multiple batches, and each wafer test item in each batch of products is tested to generate a plurality of package post-package tests An analysis method of the post-package test parameters related to the project and the package test station related to the post-package test includes: searching the database to obtain the plurality of batches of products and comparing the post-package test parameters to select the test parameter values and Correspondingly, one representative judges that the representative post-package test items are relevant; when judging the representative post-package test items is off ', it is divided into at least two first product groups based on the representative post-package products , The first qualified product group and a first unqualified product are searched for the first qualified product group in the packaging system, and the first unqualified product group is searched for the packaging station; and the first unqualified product group is judged to pass through the probability Machine. It is used to analyze the test parameter values of multiple batches of products after a plurality of machine circles are packaged in a plurality of channels. These packaged values are stored in a database with these packages and the post-packaged tests are performed. Parameter value; a representative post-package test item after package test; whether the test item is related to the package process station and the package process station is a test item to group the plurality of batches; the product group includes a first batch Machines passed by the station type; Machines passed by the process station level are higher than the first qualified product group after the package as described in the first patent application scope, among which the representative post-package test items / test parameter analysis methods are Strangulation 583406 6. Scope of patent application (k i 1 1 e d) Test items after packaging. 3. After applying the package as described in item 2 of the patent scope, test the parameter analysis method i, where the post-package test is the highest failure rate among these post-package test items. 4. The post-package test parameter analysis method described in item 2 of the patent application scope, where the post-package test item is the package with the highest average fai 1 ed rate among multiple batches of wafers. Post test project. 5. The post-package test parameter analysis method described in item 1 of the scope of the patent application, which uses a common analysis method to determine that the pass rate of the first unqualified product group is higher than the pass rate of the first qualified product group. . 6. The post-package test parameter analysis method described in item 丨 of the scope of the patent application, wherein each wafer in each batch of products undergoes an on-line quality inspection project and a sample test project related to the post-package test project. Inspection j generates an online quality inspection parameter value and a sample test parameter value. The database further stores the online quality inspection item, the sample test item, the online quality inspection parameter value, and the sample test parameter value. The test = number analysis method further includes: When Tian judges that the representative post-package test items are not related to the packaging process station, the plurality of products are divided into at least two second according to the representative post-package test items. Product group, the second product group contains a 583406 6. Scope of applying for patents ---- s-- the second qualified product group and the second unqualified product group; search the second qualified product group and the second unqualified product group from the database respectively At least one of the online quality inspection item and the sample test item related to the representative post-packaging test item, and the corresponding online quality inspection parameter value and the sample test parameter value are at least one again; ~ Statistically Analyze whether the search result of the second unqualified product group is different from the search result of the second qualified product group; § When the analysis result of the search result is no difference, stop the analysis action; When the analysis result of the search result is different To statistically analyze whether the out of spec portion of the second unqualified product group is significantly different from the out of spec portion of the second qualified product group according to a preset specification; when based on the preset, When there is a significant difference in the analysis results of the specifications, it is judged that the reason for the high value of the representative post-package test parameters is that the search is related to that The test results of at least one of the online quality inspection item and the sample test item related to the apparent post-package test item are shifted; when there is no significant difference in the analysis result according to the preset specification, a narrowed specification (narrowed spec) Statistically analyze the second unqualified product, whether the non-compliant reduced specification (out 〇f narr〇wed spec) part is significantly different from the non-compliant reduced specification part of the second qualified product group; When the analysis result is no obvious difference, stop the analysis Page 24 583406 VI. Application for patent scope action; and when there is a significant difference in the analysis results based on the narrowed specifications, it is judged that the representative post-packaged test parameter value is too high from the searched and the representative Caused by at least one of the online quality inspection item and the sample test item related to the post-package test items. 7. The post-package test parameter analysis method described in item 6 of the scope of the patent application, which determines the mean and variation of the search results of the second unqualified product group and the search results of the second qualified product group Whether there is a difference in the number (variance) to analyze whether there is a difference between the search result of the second non-conforming product group and the search result of the second qualified product group. 8. The post-package test parameter analysis method as described in item 6 of the scope of the patent application, wherein the delta specification has a preset specification upper limit (UPS pec limit) and a preset specification lower limit (i0w Spe (;). 9. The post-package test parameter analysis method described in item 8 of the scope of the patent application, which compares the number of product batches in the second unqualified product group that exceeds the upper limit of the preset specification and the total number of the second unqualified product group. Is the ratio of the number of product batches different from the ratio of the number of product batches in the second qualified product group that exceeds the preset upper limit to the total number of product batches in the second qualified product group. The post-package test parameter analysis method described in item 8, which compares the second unqualified product group below the preset lower specification limit Page 25 583406 VI. The ratio of the number of batches of products in the scope of patent application to the total number of batches of the second unqualified product group and the number of batches of products in the second qualified product group that are lower than the lower limit of the preset specification account for the Are there any differences in the proportion of total product batches for the first qualified product group? 11. The post-package test parameter analysis method described in item 8 of the scope of patent application, wherein the far-reduced specification has an up narrowed spec limit and an i0w narrowed spec 1 1 m 11 'The reduced specification upper limit is smaller than the preset specification upper limit, and the reduced specification lower limit is greater than the preset specification lower limit. 1 2. The post-package test parameter analysis method described in item 丨 丨 of the scope of patent application, wherein the range of the preset specification is six times a standard deviation, and the upper limit of the reduced specification is twice the upper limit of the preset specification. The standard deviation, the standard deviation of the reduced specification lower limit and the preset specification upper limit is doubled. 1 3. The post-package test parameter analysis method described in item No. 丨 0 of the scope of patent application, which compares the number of batches of products in the second unqualified product group that exceed the reduced specification limit to the second unqualified product group. Is there a difference between the ratio of the total product batches and the ratio of the number of product batches in the second qualified product group that exceeds the reduced specification limit to the total product batches of the second qualified product group. 14. The post-package test parameter analysis method described in item 10 of the scope of the patent application, which compares the number of batches of products in the second unqualified product group that are lower than the reduced specification lower limit to the second unqualified product group. The proportion of the total product batches, Page 26 583406 6. Scope of patent application Whether there is a difference between the ratio of the number of batches of products in the second qualified product group that is lower than the lower specification limit to the total number of product batches in the second qualified product group. 15. The post-package test parameter analysis method as described in item 6 of the scope of patent application, wherein when the analysis result according to the preset specification is significantly different, a CDF plot s output is used to find the search result and the representative The test results of at least one of the online quality inspection item and the sample test item related to the post-package test items. 16. The post-package test parameter analysis method described in item 6 of the scope of patent application, wherein when there is a significant difference in the analysis result according to the reduced specification, a CDF p 〇ts output is used to find the search result and the device. The test results of at least one of the online quality inspection item and the sample test item related to a representative post-package test item. Page 27